Road Traffic Conditions Classification Based on Multilevel Filtering of Image Content Using Convolutional Neural Networks

Abstract

Classification of traffic conditions is a vital task for determining traffic control strategies in ITS. Systematic assessment of the volume of traffic enables appropriate changes of control measures for directing traffic streams to reach set goals of performance. Video traffic monitoring is a suitable and convenient source of traffic data. The paper presents a method of classification of road traffic conditions based on video surveillance data. Convolutional neural network is used to classify the video content and establish measures of congestion of the observed traffic. Four levels of traffic conditions are distinguished which correspond to LOS categories. The network is validated using video data from several traffic observation sites. The trained CNN is capable of processing video data for systematic use by subsystems of ITS responsible for traffic management. The results of classification are compared with neural network based classifiers: a MLP (multi layer perceptron) and a DLN (deep learning network) with autoencoders. The proposed method is more accurate and less sensitive to the quality of video data.